What is the difference ?

IIRC the "CUDA" refers to the older toolkit version. the 2.3 version is an update with improved performance. at Seti the CUDA gives you WU's that use the 6.08 app where the CUDA 2.3 use 6.09

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

Are these being handed out automatically? Do I need to download an test version of seti or something? I've been trying to find these units, but so far no luck. 2 of the new rigs have gtx 465 overclocked cards and thought I would get some of these units, just not sure from the outage if they are being produced or not. Any info would be helpfull.

---

It's mostly decided by the videocard drivers you install which program Seti sends you. Since you're using Fermi GPUs, you're even one step further and should get the cuda_fermi application, as neither CUDA nor CUDA2.3 will work correctly on Fermi style GPUs. Blame Nvidia for their weird sense of humor.

Newer CUDA applications built for Fermi GPUs can run on older non-Fermi GPUs, but trying to run older non-Fermi CUDA applications on Fermi GPUs will cause errors.

---

Thank you for clarifying that, dang nvidia. I was really lookin foward to using these new systems for some of the newer units.

---

Actually the work units are all the same as far as 6.03, 6.08, 6.09 or 6.10. They are all multibeam work units to SETI and produced by SETI in the same way.

The difference lies in the science application software and the hardware used to process them (GPU or CPU) and that determines which number gets assigned to you and the work unit. That number is assigned by SETI servers after looking at your hardware and installed software.

---

Boinc....Boinc....Boinc....Boinc....

So how do you tell what the requirements are for your systems for the newer apps? Or does the server automatically start sending out units to test the system? or just simply download the toolkit and then reinstall lunatics app?

---

So how do you tell what the requirements are for your systems for the newer apps? Or does the server automatically start sending out units to test the system? or just simply download the toolkit and then reinstall lunatics app?

I need an answer to this, too.

I'm running CUDA 23 because that's what I was sent. But I'm getting way too many errors (no overclocking, no optimizations, etc) for my comfort. I've been away and left my quad-core AMD running 50% CPUs and CUDA (for heat's sake while I was away).

I'd say I have a video card problem (like low voltage or something), but I have also had some CPU errors AND the CUDA 23 errors are on both a GT 240 and GT 9800, both low-wattage cards, in two different computers.

Are there just more CUDA errors or do I have a problem?

---

You should look at the task details for the reason of the errors.

The only erroneous task remaining in the list of the GT 240 ended with "SETI@home error -12 Unknown error", which is no real error but "cudaAcc_find_triplets erroneously found a triplet twice in find_triplets_kernel".

All the tasks I checked for the 9800 GT ended in "Maximum elapsed time exceeded", which is a server scheduler problem.

There are plenty of threads about both "errors" on the number crunching forum. Try an advanced forum search for "-12" or "-177".

Gruß,
Gundolf
PS: The CUDA application you receive depends on the drivers you've installed.

I have the latest drivers installed 260.99 from Nvidia using GTX 460 card, still got errors..:(
Some one any comment on this?

Kind regards,

Mark

---

To boldly go where no human has gone before!

Not without you giving a clue about what errors you see. Since you have your computer hidden from view, we can't check into what error you see.

---